On 01/27/2013 11:10 AM, David Woodhouse wrote: > On Sun, 2013-01-27 at 11:02 -0800, H. Peter Anvin wrote: >> This is intentionally a protocol only patch, which may be possible to >> push into 3.8 as an urgent patch. David, if I understand our >> discussions right it might be better to not export XLF_EFI_HANDOFF_32 >> at this time? > > That won't stop broken bootloaders from jumping to $handover_offset > anyway. Anyone who was daft enough to implement EFI boot stub in their > bootloader *despite* its obvious brokenness, rather than insisting on > fixing it before it could be considered usable, will still be jumping > into hyperspace. I have little sympathy for them, but I'm told we need > to care. At the moment I'm concerned with a new updated bootloader, which knows the 2.12 protocol, encountering an old kernel. > It'd probably be better to just mark CONFIG_EFI_STUB as 'depends on > BROKEN' for 32-bit. > > Or, and perhaps this is heresy, merge the patches which bloody fix it? I think we can probably do that, since it doesn't affect anything non-broken at this point. I'm sorting out what can be done for 3.8 vs 3.9 at this point. Anyway, as you can tell I'm spending this weekend working for a reason. It turns out the patch I sent out doesn't actually build. Here is an updated patch. Can I get your ack for this so I can do the appropriate hacks to your and Yinghai's patchsets? -hpa -- H. Peter Anvin, Intel Open Source Technology Center I work for Intel. I don't speak on their behalf.